Am 23.02.2011 16:47, schrieb Alan Stern: > On Wed, 23 Feb 2011, Flynn Marquardt wrote: > >>>> Hi Alan, >>>> >>>> some years ago I reported a problem consuming and now I ask with another >>>> problem I have. >>>> If you are not the right person, please tell me who I should ask instead. >>>> >>>> My problem: >>>> >>>> I use synce-hal (http://www.synce.org/moin/) to connect to Windows CE >>>> devices through usb. >>>> All works fine with older devices, that only have usb full-speed and on >>>> the linux host e.g. the uhci_hcd >>>> module is used: connection can be established and used. >>>> >>>> Newer devices have usb high speed ports and linux uses the ehci_hcd >>>> module and the connection cannot be >>>> "fully" established (connection is little bit tricky, details are here >>>> http://www.synce.org/moin/TechnicalInformation/InitalConnection). >>>> The needed pppd can be started and the ppp-connection can be >>>> established, but the device cannot be 'pinged', >>>> the Windows CE is waiting for some magic bytes, that do not arrive and >>>> gets a timeout. >>>> After unloading ehci_hcd (forcing full speed) these devices work fine too. >>>> So this really a ehci related problem. >>>> >>>> Can this be fixed, so that synce also works with high speed? >>>> In Windows all works fine with these devices using high speed. >>>> I'll do any debugging, that is needed, sending usb-debugfs, usb-monitor >>>> output, ... >>> What version of the Linux kernel are you using? >>> >> I have this problem with different current versions on different machines: >> - 2.6.38-rc6 >> - 2.6.37.1 >> - 2.6.36.4 >> >>> It would help to see usbmon traces for both a working full-speed >>> connection and a non-working high-speed connection. >>> >> The output of usbmon is attached for kernel version 2.6.37.1: >> - 6.uhci.mon.out: successful connect and status query (synce-pstatus) >> with ehci module unloaded >> - 4.ehci.mon.out: failed connect with ehci module loaded >> >> The usbmon output contains only the traffic belonging to the device, e.g.: >> >> cat /sys/kernel/debug/usb/usbmon/4u > 4.ehci.mon.out > Okay, good. The two usbmon traces are very similar, right up until the > end. > > At the end, the EHCI trace got an EOVERFLOW error. The synce-hal > program asked for a 256-byte transfer and the device sent more than 256 > bytes. On a full-speed connection this doesn't matter, but on a > high-speed connection it does. > > The solution is to fix synce-hal. It should request a 512-byte or > 1024-byte transfer, not a 256-byte transfer. > > You should report this bug to the Synce project. > I tried to fix it myself and searched for 256-byte buffer and did not find any ... But I fixed it anyway: the problem is the bulk_in_size in drivers/usb/serial/ipaq.c: .bulk_in_size = 256, I increased these values to 512: .bulk_in_size = 512, and it works now and is faster! Some other serial usb drivers also have this increased value, so it should be safe. diff is attached. I appreciate your help Alan! Flynn
--- drivers/usb/serial/ipaq.c-orig 2011-02-23 19:01:44.316173437 +0100 +++ drivers/usb/serial/ipaq.c 2011-02-23 19:07:23.525877261 +0100 @@ -557,7 +557,7 @@ .description = "PocketPC PDA", .usb_driver = &ipaq_driver, .id_table = ipaq_id_table, - .bulk_in_size = 256, + .bulk_in_size = 512, .bulk_out_size = 256, .open = ipaq_open, .attach = ipaq_startup,
Attachment:
signature.asc
Description: OpenPGP digital signature
